Website Legal

Website Legal

Generate jurisdiction-specific legal pages from reviewable boilerplate templates. The skill never invents legal text — all wording comes from the template bundles in references/jurisdictions/. Claude only fills {{placeholders}} with facts captured from the user.

Disclaimer

The generated pages are templates, not legal advice. Always print the disclaimer in the final summary: "Diese Texte sind Vorlagen. Vor der Veröffentlichung durch eine Anwältin oder einen Anwalt prüfen lassen." Adapt the disclaimer to project language.

Prerequisites

A website-project.json must exist in the working directory. If missing, redirect to website-setup. A website-plan.json is not required — website-legal can run before or after website-plan. If the plan does not yet exist, the skill writes the legal page entries to a queue file (legal-pages.json) that website-plan will pick up.

Workflow

1. Load Project

Read website-project.json. Extract:

• language (drives all user-facing prompts and template language; default de)
• company (name, tagline, contact)
• legal_config (may be missing, partial, or complete)

If legal_config is missing, create an empty object.

2. Determine Jurisdiction

If legal_config.jurisdiction is not set, ask the user via AskUserQuestion:

"In welcher Rechtsordnung wird die Website veröffentlicht?"

Options:

• Deutschland (DE) — TMG §5, DSGVO, Telemediengesetz
• Österreich (AT) — ECG §5, DSG, DSGVO
• Schweiz (CH) — revDSG, freiwillige Anbieterkennzeichnung
• EU (übrige Mitgliedstaaten) — DSGVO, ePrivacy, generischer Legal Notice

Persist the choice to legal_config.jurisdiction.

3. Capture Legal Entity Facts

Read references/legal-config-schema.md to learn the full field set, then check which fields are still missing for the chosen jurisdiction (see the requirement matrix in that file).

Ask only for missing fields. Adapt the questions to project language. Group questions logically — capture the legal entity in one round, then the responsible person, then VAT/registration in a third round so the user is not overwhelmed.

Required-by-jurisdiction matrix (high level — see references/legal-config-schema.md for exact rules):

Persist captured fields immediately to website-project.json → legal_config after each round, so a long capture session is never lost.

4. Validate Required Fields

Before generating templates, run a completeness check against the matrix. If required fields for the chosen jurisdiction are still missing, list them and ask the user one more time. If the user explicitly says "use placeholder for now", insert a literal «TODO: <field>» marker — never invent values.

5. Select Template Bundle

Pick the bundle directory from ${CLAUDE_PLUGIN_ROOT}/skills/website-legal/references/jurisdictions/:

Each bundle contains three files: an imprint/legal-notice template, a privacy template, and a cookies template, all with the .md.tmpl extension.

6. Render Templates

For each template file in the bundle:

1. Read the template content with the Read tool.
2. Substitute every {{placeholder}} token with the corresponding value from legal_config, company, or the current date. Supported placeholder paths are documented in references/placeholder-schema.md.
3. If a placeholder has no value, insert the literal string «TODO: <placeholder>» so the gap is visible in the rendered HTML.
4. Replace {{generated_at}} with the current date in ISO format and {{generated_at_human}} with the localized form (e.g. "8. April 2026" for de).
5. Write the result to content/legal/{slug}.md in the project directory. Create the content/legal/ directory if it does not exist.

The output is markdown, not HTML. The existing page-generator agent renders these markdown files to HTML during website-build using the legal page templates from ${CLAUDE_PLUGIN_ROOT}/libraries/legal-pages.md.

7. Patch website-plan.json (if it exists)

If website-plan.json exists in the project directory:

1. Read it.
2. Remove any existing legal page entries (page IDs starting with legal-) so re-runs are idempotent.
3. Append three new page entries — one per generated legal page — using the structure from ${CLAUDE_PLUGIN_ROOT}/libraries/EXAMPLE_WEBSITE_PLAN.md (legal section). Each entry has:
   • id: legal-imprint, legal-privacy, legal-cookies
   • type: legal-imprint | legal-privacy | legal-cookies
   • slug: pages/impressum | pages/datenschutz | pages/cookies (or legal-notice/privacy-policy for EU)
   • title: localized page title — site title
   • meta_description: short SEO description (e.g. "Impressum gemäß § 5 TMG")
   • source_files: ["content/legal/{slug}.md"]
   • sections: ["legal-header", "legal-body"]
   • footer_only: true — exclude from primary nav, include in footer legal column
4. Add or replace the top-level legal_links array. Labels and slugs come from the jurisdiction bundle and the project language — never hardcode German. The DE example below is illustrative only; for eu the slugs become legal-notice / privacy-policy and the labels are English:

   [
     { "label": "Impressum",   "href": "/pages/impressum.html" },
     { "label": "Datenschutz", "href": "/pages/datenschutz.html" },
     { "label": "Cookies",     "href": "/pages/cookies.html" }
   ]
   

5. Write the updated plan back atomically.

If website-plan.json does not yet exist, write the entries and the legal_links array to legal-pages.json in the project directory. The website-plan skill will merge legal-pages.json into the plan it generates (see the page-type-registry for details).

8. Print Summary

Render the summary in the project language. The German example below is the shape, not a literal template — translate the labels when language differs.

Rechtliche Seiten erstellt für Rechtsordnung: {jurisdiction_name}

  Generierte Seiten:
    ✓ content/legal/impressum.md     ({n} Wörter)
    ✓ content/legal/datenschutz.md   ({n} Wörter)
    ✓ content/legal/cookies.md       ({n} Wörter)

  Konfiguration:
    Rechtsträger:    {legal_entity.legal_name}
    Anschrift:       {legal_entity.address.city}
    Verantwortlich:  {responsible_person.name}
    USt-IdNr.:       {legal_entity.vat_id}

  Offene Platzhalter: {n}    ← bitte vor Veröffentlichung füllen
  Plan aktualisiert:  ja|nein (legal-pages.json zwischengespeichert)

⚠ Diese Texte sind Vorlagen. Vor der Veröffentlichung durch eine Anwältin oder
   einen Anwalt prüfen lassen.

Nächster Schritt: /website-build — Seiten generieren (oder /website-plan, falls
                  noch nicht ausgeführt).

If any «TODO: ...» placeholders are still present, list them explicitly so the user can fix them before building.

Re-running the Skill

The skill is idempotent. Re-running it:

• Re-reads legal_config from website-project.json
• Asks only for fields that are still missing
• Overwrites content/legal/*.md with freshly rendered templates
• Replaces existing legal-* page entries in website-plan.json (no duplication)

If the user wants to switch jurisdictions, ask them to confirm — switching deletes the old templates and writes new ones. Old captured facts in legal_config are preserved where the new jurisdiction needs the same fields.

Output Language

All user-facing text and template output use the project language field. Default de. Each jurisdiction bundle currently ships with German templates (DE/AT/CH) and an English template (EU). When the project language differs from the bundle language, render the templates as-is and warn the user.

Reference Files

• ${CLAUDE_PLUGIN_ROOT}/skills/website-legal/references/legal-config-schema.md — full schema and per-jurisdiction requirement rules
• ${CLAUDE_PLUGIN_ROOT}/skills/website-legal/references/placeholder-schema.md — list of supported {{placeholders}} and their source paths
• ${CLAUDE_PLUGIN_ROOT}/skills/website-legal/references/jurisdictions/{de,at,ch,eu}/ — boilerplate templates
• ${CLAUDE_PLUGIN_ROOT}/libraries/legal-pages.md — HTML/CSS pattern for legal page rendering
• ${CLAUDE_PLUGIN_ROOT}/libraries/EXAMPLE_WEBSITE_PLAN.md — reference for the legal page entries and legal_links array